May 19, 2020 | Written by: Katia Moskvitch
Share this post:
Researchers globally have been using the world’s fastest computers thanks to the COVID-19 HPC Consortium for nearly two months now – but there is still supercomputing capacity, and the partnership is calling for more proposals. “There is real hunger on the free resource providers side for good projects,” said Jim Brase, Program Leader at Lawrence Livermore National Laboratory, at a recent webinar that united the leadership of the initiative. “We’d like more proposals to be coming in.”
At the moment, there are 38 members of the Consortium and 51 active projects that have been matched with a supercomputer. “While many of the current members are competitors, in the context of COVID-19, the world united,” said Dario Gil, the Director of IBM Research. “It has been a pleasure to work together with my colleagues from Google, Microsoft, Amazon, HP and many others, to bring the best of our respective institutions, working with our colleagues from academia and the many participating institutions from DOE to NSF to NASA – it has been extraordinary. Joining forces will only make our respective institutions stronger, and it’s a sure win for society.”
Many proposals have so far been from the US, although eleven have been submitted from other countries. Some are already starting to produce results – merely weeks after beginning to crunch the numbers on supercomputers, said Paul Dabbar, Under Secretary for Science at the Department of Energy.
For instance, one team at the University of Michigan is using XSEDE Jetstream supercomputer with AI capabilities to try and repurpose existing drugs to tackle COVID-19. The researchers have built deep learning models to screen over 1,600 current FDA-approved drugs and more than 5,000 experimental drugs in the drug bank. Algorithms will then use these molecules to find existing drugs that might be best suited to help provide therapeutics for the disease, said Dabbar.
Another notable project, he added, is COVID Moonshot, by a British chemistry startup PostEra that uses AWS supercomputer. These researchers are looking for antiviral drug candidates to fight the virus. In two months, they have identified over 60 experimentally confirmed fragments that effectively target a key protein associated with COVID-19 – having relied on crowdsourcing to submit potential drug designs. The team has received more than 2000 submissions worldwide and has used the supercomputer to quickly figure out which ones were viable. “Just this week, PostEra has developed chemical synthesis pathways for 400 promising compounds, and 200 of those have been manufactured and are currently in testing,” said Dabbar.
There are also projects analyzing patient trajectories and outcomes, for instance, looking for links to genomics of a patient and categorizing patients on expected outcomes. Another project, from a team at Duke University, is looking into multi use ventilators; the researchers have developed a device that safely splits ventilators between two or more patients.
Another group, at Universidade Federal de Santa Catarina, has designed a mobile app and a platform for anonymous contact tracing. “The patient area is where we’d like to see more proposals coming in, we’d really like to encourage it,” said Brase. “I think there are significant computational aspects where these resources could help.”
Brase also highlighted a project by a group in Michigan State University that is developing the first high resolution structure and dynamics prediction for membrane proteins of the virus, running on the San Diego supercomputing center.
Another team that has just released results is from MIT. These researchers have developed a new model for multiple contact tracing using eye directional analysis of the data. It works by using contact tracing in both directions on the contact chain, and the scientists have obtained significantly improved performance. They have just published a paper on MedArXiv.
Gearing for the post-COVID world
The speed with which the HPC Consortium has been created is unprecedented, said Dabbar, “nothing like this has been done since World War Two, with public and private entities coming together in such short time. The access to our nation’s computing resources is essential to respond as quickly as the crisis merits, and pooling resources has helped researchers to understand how the disease spreads in communities as conditions change on the ground,” he added.
The initiative has also allowed researchers to rapidly identify targets of the virus within humans to more rapidly screen potential pharmaceutical treatments and develop a future vaccine, Dabbar said. “It is allowing researchers to start predicting how the virus may change as it spreads through the population.”
The Consortium keeps on growing, with new partners constantly joining in. One day soon, there will be a vaccine or a treatment, but the experience of setting up such a body shouldn’t go to waste. IBM Research’s Director Dario Gil, who was the initial catalyst of the Consortium and suggested the idea to the White House Chief Technology Officer Michael Kratsios in mid-March 2020, thinks that we should apply a similar model on a wider scale in the post-COVID era.
Gil calls for the creation of a new international organization he provisionally calls the Science Readiness Reserves. It would unite academia and industry and provide expertise and resources to deal with the next ‘known unknown’ – the next global emergency our world may face. It could be another virus, or a mega drought, a volcano eruption or even an asteroid strike over a city. This organization would engage in long-term planning for unexpected events and be an active science reserve that could be mobilized in times of a global threat.
“This could be a coalition of willing scientists and participants that would come together in moments of crisis like we’re doing right now,” said Gil. With one key difference: we would be prepared much, much better.